Genre as a Classification Problem

Scholars have long understood that nineteenth-century newspaper genres are deeply hybrid and amorphous, melding into one another across any ostensible boundary. This is the case particularly after the rise of the so-called “penny papers” in the 1830s. In her study of literary hoaxes in penny press journalism, Karen Roggenkamp (2005) writes, “penny papers specialized in news that could easily be confused with fiction. Through a readjustment in the public’s standard for judging newspapers, the penny press smeared preconceived borders between fact and fiction in the antebellum literary marketplace” (2). Sari Edelstein (2014) adds that this genre soup was specifically aimed at a new, urban readership: “With a heavy emphasis on scandal, gossip, and crime stories, penny papers appealed to newly urban populations looking to the press for a blend of information and amusement” (4) The dynamism of the periodical press is one of the most compelling features of the medium for literary-historical inquiry, but it also makes it difficult to grapple with questions of genre at scale.

At a conceptual level, genre relies on intellectual models for classification. Much as humanists might shy away from classification as it is operationalized in bureaucratic and computational systems, humanistic work to sort literature into categories requires its own kind of operationalization, whether implicit or explicit. In order to decide that this novel is gothic while that novel is sentimental requires that a scholar inventory those features of style, theme, character, or tone that distinguish members of the group “gothic” from members of the group “sentimental,” and to then apply those identified features to particular texts in order to determine where they lie. As Underwood (2019) notes, scholars from different disciplines focus on different features when distinguishing genres, seeing them as “communicative actions” (for rhetoricians), “communities of readers” (for sociologists), or textual patterns (for literary scholars; 34). These determinations might be complex or even ambiguous, if a given work is determined to share qualities of multiple genres, but we might imagine the task of scholars taking any of these approaches as placing individual works in relationship to others on a mental scatterplot.

In her introduction to the October 2007 special issue of PMLA, Wai Chee Dimock (2007) writes, “Far from being a neat catalog of what exists and what is to come, genres are a vexed attempt to deal with material that might or might not fit into that catalog” (1378). Dimock goes on to conceptualize genres as “open sets endlessly dissolved by their openness ... defined over and over again by new entries that are still being produced” (1379). What we call science fiction today seems quite different from what we might identify as science fiction in the late nineteenth century, though as Underwood demonstrates in Distant Horizons that distance is less extreme than we tend to imagine, and computational models are well suited to recognizing similarities where human beings might miss them. As he writes, “A problem like this [genre] requires a methodology that is cautious about ontological assumptions and patient with details.” Along with Underwood (2019), we argue in this chapter that “Predictive modeling fits the bill” because it is focused not on writing a more perfect definition of any given genre, but instead on finding “a model that can reproduce the judgements made by particular historical actors” (41). In our case, we draw on computational models to gain a better perspective on how nineteenth-century readers might have assessed the vignettes they came across in their newspapers: the ways they would have seen these texts as similar to or distinct from the other texts they were printed alongside. As we will see, computational models are quite effective at reflecting back to us readers’ ambiguities and disagreements about genre. The model, after all, only knows as much about genre as we can teach it.